The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies heavily on system tools for media manipulation. Multiple scripts (scripts/tts.py, scripts/sync_check.py, scripts/audio_check.sh, scripts/check_video.sh, and scripts/contact_sheet.sh) invoke ffmpeg and ffprobe via subprocess calls to encode video, extract audio, and validate media properties. These operations are consistent with the skill's primary function of video production.
[EXTERNAL_DOWNLOADS]: Several dependencies listed in requirements.txt involve network access:
edge-tts connects to Microsoft's speech platform (speech.platform.bing.com) to generate narration.
faster-whisper is configured to download pre-trained machine learning models from official Hugging Face repositories (Systran/faster-whisper-base.en, Systran/faster-whisper-small) for transcription and verification tasks.
[DYNAMIC_EXECUTION]: The rendering pipeline in scripts/capture.mjs utilizes dynamic execution to capture video frames:
It spawns a local HTTP server on a random loopback port (127.0.0.1) to serve the project files, with built-in path-traversal protection ensuring only the designated project directory is accessible.
It uses headless browser automation (Playwright or Puppeteer) to load the project and executes user-defined rendering code via page.evaluate, which is the intended mechanism for frame-by-frame capture.

code-video